Research map / Continual learning / Online and replay-based
HLML-SNN:Fast Continual Learning in Spiking Neural Networks Achieved via Hebbian Learning-Driven Meta-Learning
Jiangshuai Xu, Peiyun Xue, Jiacheng Song, Xuhui Huang, Qingshan Hou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add meta-learning to Hebbian learning (ablation) | HLML-SNN · split-CIFAR10 | average accuracy (%): 52.8 -> 93.1 | Improved |
| add meta-learning to Hebbian learning (ablation) | HLML-SNN · 10-split-CIFAR100 | average accuracy (%): 9.9 -> 90.2 | Improved |
| use Hebbian learning instead of standard gradient descent in inner loop (ablation) | HLML-SNN · split-CIFAR10 / 10-split-CIFAR100 | training time: 3 to 5 times faster | Improved |
| use Hebbian learning instead of standard gradient descent in inner loop (ablation) | HLML-SNN · split-CIFAR10 / 10-split-CIFAR100 | average accuracy (%) | No clear effect |
| increase SNN time steps from 4 to 10 (ablation) | HLML-SNN · 10-split-TinyImageNet (N) | average accuracy (%): 78.9 -> 80.1 | Improved |
| increase SNN time steps from 4 to 10 (ablation) | HLML-SNN · 10-split-CIFAR100 | training time (s): 9.54 -> 22.02 | Hurt |
| increase SNN time steps from 2 to 4 (ablation) | HLML-SNN · split-CIFAR10 | average accuracy (%): 91.7 -> 93.1 | Improved |
| lower meta learning rate from 1e-2 to 1e-4 (ablation) | HLML-SNN · split-CIFAR10 | average accuracy (%): 88.2 -> 93.1 | Improved |
| lower meta learning rate from 1e-4 to 1e-5 (ablation) | HLML-SNN · split-CIFAR10 | average accuracy (%): 93.1 -> 91.2 | Hurt |
| add explicit orthogonality constraints (ablation) | HLML-SNN · HLML-SNN | accuracy / training time: no gain, 4-fold training time | No clear effect |
| use HLML-SNN instead of LWF | HLML-SNN vs LWF · split-CIFAR10 | average accuracy (%): 84.51 -> 93.1 | Improved |
| use HLML-SNN instead of LWF | HLML-SNN vs LWF · split-MNIST | average accuracy (%): 99.10 -> 99.1 | No clear effect |
Closest papers by meaning
- Hebbian Learning based Orthogonal Projection for Continual Learning of Spiking Neural Networks (ICLR 2024 · similarity 0.89)
- Learning the Plasticity: Plasticity-Driven Learning Framework in Spiking Neural Networks (NeurIPS 2025 · similarity 0.89)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.88)
- Efficient Spiking Neural Networks with Sparse Selective Activation for Continual Learning (AAAI 2024 · similarity 0.87)
- ALADE-SNN: Adaptive Logit Alignment in Dynamically Expandable Spiking Neural Networks for Class Incremental Learning (AAAI 2025 · similarity 0.87)
- Self-Normalized Resets for Plasticity in Continual Learning (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
